Jump to content

Browsing the web on 98/ME in 2019 and beyond


Destro

Recommended Posts

Unfortunately PCem only emulates up to a Pentium MMX 233 and a S3 ViRGE DX/Voodoo 2 SLI, which are quite suboptimal for web browsing and D3D 7+ (and probably earlier).

The best option would be for a hypervisor to emulate a more-powerful but 9x compatible GPU, like a TNT2 or above (or even better: Radeon 9xxx or GeForce4 and above), removing the overhead of having to emulate the CPU. However such products are focused on servers, so no one has tried to pull this off. And I'd think you would need many CPU threads to get it to work acceptably as well.

Edited by win32
Link to comment
Share on other sites


1 hour ago, win32 said:

Unfortunately PCem only emulates up to a Pentium MMX 233 and a S3 ViRGE DX/Voodoo 2 SLI, which are quite suboptimal for web browsing and D3D 7+ (and probably earlier).

The best option would be for a hypervisor to emulate a more-powerful but 9x compatible GPU, like a TNT2 or above (or even better: Radeon 9xxx or GeForce4 and above), removing the overhead of having to emulate the CPU. However such products are focused on servers, so no one has tried to pull this off. And I'd think you would need many CPU threads to get it to work acceptably as well.

Thanks for this useful info! Saves me a lot of time. I can now sleep in peace, knowing that Windows 98 is dead for me (lol). And that I have to stick with Windows 2000 Pro on vmware...

Link to comment
Share on other sites

Does everyone experience a 600 pixel wide view when viewing MSFN forum now with Opera 12.02 or is it just me. I post with a full width screen and view my profile with full screen width with desktop mode 1920 x 1080. Users can choose their window width and text size to what suits but not now. 
 

Link to comment
Share on other sites

Goodmaneuver said:
> Does everyone experience a 600 pixel wide view when viewing
> MSFN forum now with Opera 12.02 or is it just me.

Oh yeah :-( Definitely not just you, and not just Opera either, but other very old browsers too, Firefox and other Geckos.
That prob first started at Feb-13, and after much struggling and fiddling have finally concluded it depends purely on engine css-capabilities. Useragents don't help at all.
And the width is just one prob now, the second annoying layout mess in such ancient browsers is the vertical stacking of author-block and comment-block now!
But funnily, that vertical prob is NOT happening in KM1.5 (Firefox2), and not in KM74 (Firefox24), only in between, in my KM1.6+1.7 (Firefox 3.5/3.6). That led me to suspect it may be caused by some 'new' html5 tags introduced in FF3.5 with just initial support, not fully yet, perhaps with other default display values at first, suspects like "aside" or "article"...
And obviously same probs in Opera12.02, both, as just tested after your post. Oh great, now must fiddle with that too, grmpf.

Anyway, meanwhile figured out workarounds for userContent.css. Not perfect, and it's adjusted to my own screen width (1024), but finally a view again that doesn't hurt they eyes, OUF!
That workaround-code is just a bit complicated and messy, meant only for selfhelp. May post it later as last resort, but hope that someone has a better solution? Perhaps should do more tests to figure out more too - wonder e.g. if the original css may work again by only exchanging aside-article with div's again, or whatever was the initial cause?

The universal prob is that this messy css with newer html-tags is now affecting LOTS of websites and forums :-( Especially breaks layout widths, usually rather too wide, needing constant scrolling left-right or killing styles completely, but often also half-screen width as here (and I always thought this were intential, being forced unto a narrow 'mobile' view!), or causing overlapping stuff, everthing needing killing all styles as workaround. And suddenly have understood why so many other forums lately have vertically stacked author-blocks, so that stupid layout is not intentional either.
So what we really could use urgently is a universal workaround for those tags!
What I also wonder: have seen that all browsers have a default css sheet deep inside omni.ja somewhere. Unlike userContent.css, that wouldn't get enforced on all sites, only deliver default values. Would it perhaps help anything to exchange this one with a newer one, containing all new tags with their default values?? Or would that break yet more? And that's probably missing OLDER tags in exchange? Anyone know if there's somewhere a finished perfect "mix" css-sheet for old browsers...?

Edited by siria
Link to comment
Share on other sites

Was hoping you had the answers siria:
https://msfn.org/board/topic/177106-running-vanilla-windows-98-in-2020/page/20/?tab=comments#comment-1177272

IMHO probably a losing battle once a browser needs to modify CSS site specific. Having used Dillo as a lightweight secondary browser for years, it's not a big issue for casual browsing but most would probably disagree. In RetroZilla, View -> Use Style -> None often needs to be selected, otherwise even links disappear from a web page. Also can't log in to this forum without setting style to none.

Using Page Up/Down keys is more efficient than excessive scrolling, due to poor page layout and huge amounts of whitespace. After a long vanilla Windows 98 session, it's always refreshing to multi-boot to a newer browser.

Link to comment
Share on other sites

Wunderbar98 said:
> IMHO probably a losing battle once a browser needs to modify CSS site specific

At the beginning not much of a prob yet. If it helps, it's great. Just occasional fiddling a bit with styles, then glad that frequent sites look neat again. But that was years ago, and meanwhile yeah, it's become a real pain. Now find myself fiddling and struggling with site-fixes almost daily, and gets harder and harder :-(

Here are some helper styles for Firefox-2 / 3.5, tried to polish my drafty code a bit.
The crucial lines are just a few, marked with "killer" ;-)
Hope the board will not mess the syntax.

 @-moz-document url-prefix(https://msfn.org/board/)  , 
url-prefix(http://web.archive.org/web/20180530181434/https://msfn.org/board/) 
{
/*=======  For Firefox3.5 + FF2? / KM1.6 */
/* KILLER since 2020-Feb-13: page only half screen width and comments BELOW author block
article=both (unhides middle titles) / aside=left  / div.ipsColumn=right :
*/
article[id^="elComment_"] {display: block !important;}
aside.ipsComment_author {display: table-cell !important;}
div.ipsColumn {display: table-cell !important; width: auto !important;}

/* Fix: crosslinks to other posts only empty rectangles */
body.unloaded #ipsEmbedLoading{position: relative !important;}

/* old KM-bug: posted images completely vanished if TLS broken */
div[data-role="commentContent"] img:not([data-emoticon]) {
min-width: 30px !important; 
min-height: 30px !important;
}

/* other bits */
a {text-decoration: underline !important;}
div[class="ipsItemControls"] {background-color: transparent !important; }


/*======= Only needed for Firefox2 / KM154 */

/* KILLER since 2020-Feb-13: page only half screen width. No chance except FIX px */
div.ipsColumn {width: 720px !important; }

/* dec 2019: FF2/KM154: Fix: title of single topic pages only 5cm wide */
body div h1 span.ipsType_break {width: 750px !important;}

/* FF2/KM154:  Fix: TAG-rows stacked vertical */ 
ul.ipsTags {display: table-row !important; }
ul.ipsTags li {display: table-cell !important;}
ul.ipsTags li a:link {color: blue !important; background-color: transparent !important;}

/* dec 2019 FF2/KM154: Fix: pagenumbers stacked vertical */ 
ul.ipsPagination {display: table-row !important; }
ul.ipsPagination li {display: table-cell !important; padding-left: 12px !important;}
li.ipsPagination_pageJump a {white-space: nowrap !important;} 

/* dec 2019 FF2/KM154: Fix: page Numbers white on white etc */ 
li.ipsPagination_first a, 
li.ipsPagination_page a, 
li.ipsPagination_last a, 
li.ipsPagination_next a, 
li.ipsPagination_prev a, 
li.ipsPagination_pageJump a {background-color: transparent !important; } 
ul.ipsPagination {background-color:  transparent !important; } 
li.ipsPagination_active {background-color:  peru !important; } 
}

/* Firefox2 / KM154: Ric: Recent post list too narrow, and titles only 5cm */
@-moz-document url-prefix(https://msfn.org/board/discover) {
h2 span {
width: 800px !important; 
padding: 4px !important; 
font-weight: bold !important; 
background-color: #ffee66 !important;
}
}


Link to comment
Share on other sites

Thanks for the code siria, looks like you spent lots of time on it. Just paste it into userContent.css jumper.

By default there is no userContent.css file, just userContent-example.css. Copy userContent-example.css to userContent.css and paste siria's code.

In RetroZilla v2.2 the pathway is:
C:\WINDOWS\Application Data\RetroZilla\Profiles\default\your_profile\chrome

With the tweaks was able to log in without changing to style none. However, can't tell i'm logged in. Still need to change back to style none to see my login stuff (notifications, settings, etc). First impression is somewhat better but still broken compared to a modern browser. Will test a bit more.

What about a style that's essentially 'None' but just removes excessive whitespace?

Link to comment
Share on other sites

On 2/2/2020 at 6:26 PM, Goodmaneuver said:

I do not want to rush KernelEx. Mozilla 9 will run with 4.5.120=4.5.2 (final) standard settings. My previous posted settings are recommended for best performance. Win95 setting on CometBird works on one build. The text indices on some places do not allocate the font with Mozilla. This is not a resource problem, I think it is not reading the registry in certain cases like that of Vuze and WinWord97 if Win95/98/ME setting not chosen and Riched20 14.0.7155.5000 is used. The text is not displayed on the "save as" layered tabs much like the browser. JavaScript is used in the OS and for Mozilla's main engine. It is a speedier method of transposing WEB pages. See speed of the Java only built browsers above, they are much slower.

wonder if you still use CometBird, if so, you may test my new NSS (with hex-patched xul.dll making nssutil3/ssl3/smime3 redirect back to nss3) to see if TLS 1.2 and newer cipher suites are working:

http://o.rths.ml/gpc/files1.rt/CometBird_11.0-new-nss.7z

Edited by roytam1
Link to comment
Share on other sites

@roytam1, your NSS files work fine with CometBird 11 but unfortunately that version of this browser is too problematic for me (Win ME + KernelEx) on many sites, total browser freeze on Google search, rendering area freeze on Yahoo search, weird characters displayed on other sites, etc..., none that has to do with your updated files.

I tried your files on an older version of CometBird which doesn't have these issues but they aren't compatible, browser crash on start in mozcrt19.dll.

Link to comment
Share on other sites

Thank you Roytam1. Original testing done with fully original CometBird11 with all your 11.0-NEW-NSS files. Later testing with my grepref.js + rzbrowser-tls12 + original XUL.

The main difference I have discovered about Mozilla 11 compared to Mozilla 9, is that 11 does not use any virtual memory where as 9 does use virtual memory. I say Mozilla is because I have tried FireFox SeaMonkey and CometBird and the XUL behaves with the same effect. I do use your files to upgrade the 9.01 Mozilla browser. I have 1GB of HDD cache set now in the grepref.zip on page 8 this topic. I use WinMe with 500MB of virtual file cache which I believe is the commit charge but I need correcting if wrong. When a fairly complex page is loaded with XUL 9, the commit charge is increased toward 100%/500MB then the HDD is being worked then when the page is finished loading, the Commit Charge, (from process explorer v10.2), then drops to zero. With XUL 11 there is no commit charge used and if web page does not change ie no ads for example, then Mozilla 11 can work and your set of files do work the TLS1.2 sites but after a period of time with XUL 11 (and as mentioned dynamic sites I think), the processor goes 100% and does not recover. Your XUL mod does stop the error in PLDS4.dll - (in my case). I have overcome this error before but the solution eludes me at the moment, it was a choice of file. If using original XUL 11, the complete swap of your upgraded rzbrowser-tls12 files work but with expected XUL 11 behavior as mentioned. Is there an advantage with the v26 files over v21.4? The NSS file versions you have of v49 work CometBird 9 but the URLs do not attempt to load. The 11 new NSS files as a whole unit excluding XUL do not work with CometBird 9.
 

Edited by Goodmaneuver
I have had to delete the last paragraph as it is not working out. Please ignor if previously read.
Link to comment
Share on other sites

XUL10 or greater are unacceptable as the virtual memory is not being used as mentioned before and if virtual memory is not used, that is the vcache, then the swap file will not get initiated from the vcache. Take for instance with the TLS2 up grade and visit GameFront, all of my RAM gets used up and then the HDD gets used with CometBird9 or 11. I have effectively 1GB of free RAM before CometBird is run. (Total 1.5GB, 200MB is used up on a RAM drive but not in use). If visiting a site you may be lucky first time with CometBird11 but it will soon go to 100% CPU usage and not recover. Once this happens the chances of getting through to the same position in a web site is worse. To speed up XUL9 to be responsive as 11, perhaps 13, is to use XPCOM.dll from FF13, this will speed responsiveness until the HDD is needed. Best to make use MSVCR80 or 90 to match other browser modules. As mentioned on page 8 other SM10 (SeaMonkey) files can be used as well. If you use my grepref.js then this should speed up performance too but I think the browser agent may be causing timing issues with MSN portal. An error page is sent when pages are nearly fully loaded. It is a much more difficult load compared to Opera or IE6. I did not notice any problem happening before nor with original grepref.js. It is the only site that I find that has a problem like this.

Just confirmed XP FF31 user agent string possibly a bit ambitious. Does work on most sites including MSN and YouTube but timing is what I think is taken into account at MSN. It seems to fail when skipping thought the servers at speed though, but I think throughput is still what the test is or there is not enough choking - likening to torrent transfers. I have crippled the browser in certain respects but does perform better but my NIC card can only go at 3.7MB / sec so tuning browser will depend on ethernet and internet speed. MSN home page loads fast enough then click on something of interest before package data failure message occurs. (This page is not available at the moment) Using this UA makes MSN portal a much more complex web page with many ads in comparison to older browser settings like Opera12 or IE6. If choosing a moderner OS as UA then even more throughput is expected but web pages can load faster. I tried to make a minimum of buffer and connections to streamline memory usage.

Edited by Goodmaneuver
Added better description of MSN message in brackets.
Link to comment
Share on other sites

@roytam1, could you make these files for CometBird 8 too?

Your files for CometBird 9 work fine but unfortunately v9 is as bad as v11 for me (nothing to do with your files) whereas v8 works without any of those horrible site-dependent freeze issues.

Thanks in advance.

Link to comment
Share on other sites

The only difference between the CB9 and FF10 files is that CB9 ones are linked to msvcr90 and FF10 ones are linked to msvcr80. They should work with all FF versions. Or you could try rzbrowser-tls12-20200127.7z ones which are linked to msvcrt. My grepref.js might help. FF3 could really only cache to the HDD up to 100MB or it would get into trouble. I have turned browser cache to the HDD off but the HDD will still be used a lot. I said I upped the size to 1GB that is if it is enabled. The UI can set size which I set 1GB so I did it as default. Forgot I had it disabled. When experimenting the 1GB worked well until it filled up then froze. The extra speed comes from pipelining. Originally pipelining was disabled. Original pipelining settings were OK though and tuning made an improvement but I found someone else's method and I stuck with theirs.

Edited by Goodmaneuver
Made more sense
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...